Skip to content

Latest commit

 

History

History
66 lines (54 loc) · 3.11 KB

README.md

File metadata and controls

66 lines (54 loc) · 3.11 KB

Multi-Modal-Retinal-Fundus-Scan-Analysis

A multi-modal approach for multiclass ophthalmic disease classification for fundus scan images.

Overview

The Eye Doctor System is an application designed to assist in diagnosing eye diseases using fundus images. It consists of a Flask-based backend that receives patient information including age, sex, and left/right fundus scans. This data is then processed by a deep learning model to predict the confidence percentage of various eye diseases. The predictions are sent to the front-end for display, where users have the option to view a diagnostic report generated by an AI Large Language Model.

System Components

Backend (Flask App)

  • Receives patient information and fundus images
  • Processes data using a deep learning model
  • Generates confidence percentages for eye diseases

Frontend (HTML/CSS/JavaScript)

  • Displays predictions and diagnostic reports
  • Provides user interface for interaction with the system

Demonstration of Front End

Deep Learning Model

  • Processes input data to predict eye disease probabilities
  • Utilizing late fusion technique to combine the patient demographich features with the Fundus scan features

Illustration of the Model Architecture

AI Large Language Model

  • Generates diagnostic reports based on patient information and disease predictions
  • With carefull promt engineering, focus of the ChatBot is driven towards the diseases with high confidence score
  • A diagnostic report is generated while considering the patient's age, gender, etc.
  • The report consists of explaination, possile cures and next step to be taken in the right direction

Usage

  1. Start the Flask backend by running python app.py.
  2. Open the frontend interface (index.html) in a web browser.
  3. Enter patient details and upload left/right fundus scans.
  4. View the predicted confidence percentages for various eye diseases.
  5. Optionally, request a diagnostic report from the AI Large Language Model.

Obtaining and Training the Model

  1. Obtain the OIA-ODIR dataset from [https://drive.google.com/file/d/1-7DO1jJFC_4W0hc2CaonlLe595M4eDOh/view].
  2. Follow the instructions provided in the provided Notebook[FUNDUS-DEEP-NET-AUGMENTED.ipynb] file to train the model.

Obtaining and Training the Model

  1. Obtain the OIA-ODIR dataset from [https://drive.google.com/file/d/1-7DO1jJFC_4W0hc2CaonlLe595M4eDOh/view].
  2. Follow the instructions provided in the provided Jupyter NoteBook file to train the model.

Installation

  1. Clone the repository to your local machine: 
    git clone https://github.com/your-username/eye-doctor.git
  2. Install the necessary dependencies: 
    pip install -r requirements.txt
  3. Start the Flask backend: 
    python app.py
  4. Open the frontend interface (index.html) in a web browser.

Dependencies

  • Flask
  • Deep learning framework (e.g., TensorFlow, PyTorch)
  • An AI Large Language Model library (e.g., OpenAI GPT)

License

This project is licensed under the MIT License - see the LICENSE file for details.